This interview first appeared on HBI’s blog.
Helfie AI is an Australian start-up that is building what its founder, George Tomeski, describes as “a whole new operating system for human health”. This is not simply hyperbole or PR-speak. What Helfie is building is indeed groundbreaking.
Helfie is an app that anyone can download onto their phone that gives them access to a personalisable AI-driven system for early disease detection and prevention. Using just the sensors available on your phone — image, video, audio — it can check whether you are likely to have a range of common health conditions, simply by taking a picture of a part of your body or by taking an audio recording of a cough, for instance. Currently it covers about 30 conditions, but Tomeski says that the current R&D they are doing suggests they will soon be able to do such tests for 90% of all known conditions.
And it’s very accessible cost-wise: each check costs about $0.10–0.20.
“Our body tells us a lot,” Tomeski says. “We want to operate at that computational boundary where we can understand what might be going on, and give you an insight as early as possible. If we operate in that window, you can change outcomes for everyone: for individuals themselves most importantly, but also for the health system.”
George Tomeski, CEO, Helfie AI
When asked whether it could be described as a (new type of) symptom checker, Tomeski demurs, saying: “The way I describe it is we’re building a personal health algorithm for every human. That algorithm can do various symptom checks, but it can also proactively engage with you in and around your health.
“We do a general health checkup when you sign up; you just need to scan your face by taking a selfie, we measure your blood pressure, your heart rate, your respiratory rate, your heart rate variability, your stress level, your BMI. If you say to Helfie, ‘I’m not feeling well’, the first thing it will do is ask to check your vital signs. And then it engages you in a conversation around why you aren’t feeling well. Another use case: if you’ve got diabetes, you can just hover your phone over your food and Helfie will say ‘yes’ or ‘no’.
“It’ll ask you stuff. If you’re coughing, it can say ‘this looks like COVID’ or ‘this looks like TB’ or ‘this looks like a COPD signature’. It’ll then do symptomatic checks. We had a case of a 44 year old woman with a hurt knee. The algorithm had assessed that she was in early stage menopause; it had nothing to do with arthritis. Because it did a general health check and connected a bunch of other bits of information it was able to determine that.
“It’s what I call ‘intermodal communication’. It is effectively a whole new operating system for human health — one that is dedicated to early detection, prevention and optimising health, not for anybody else, but for you.”
Tomeski says the app has received CE Class IIB medical device certification in the EU, although it doesn’t actually need this as it is not providing a diagnosis (in the same way ChatGPT or Google don’t need to be regulated as medical devices).
“We steer away from the word ‘diagnosing’ for a couple of reasons. One is that when a doctor puts a stethoscope on your chest and says you’ve got to go get some test, they aren’t doing a diagnosis. They’re telling you there could be something wrong and you need to do an X-ray which will give you a more definitive diagnosis,” he says.
“Having said that, we can give you a pretty accurate indication of what you’re dealing with. For example, in the case of skin conditions, we’ll tell you that there is a high risk for melanoma, and tell you how confident the algorithm is about that. It won’t say ‘you’ve got cancer’, because it’s not until someone does a biopsy that that can be determined.
“What it does tell you is ‘we’ve observed properties in this particular image that are indicative of X’, based on the data sets and the training of the algorithms. When we take a photo of your stool, for example, we tell you that this discoloration is consistent with signs of bowel cancer, and we can eliminate diet as a possible cause through conversation with the algorithm. And that’s a prompt for you to act on. It’s not to say you’ve got bowel cancer, but it’s to say that we have sufficient evidence now to say that you’ve got to act. Currently we can’t look at blood in stool with the algorithm, but probably by February next year we will be able to, and give the user the highest possible fidelity in terms of what they might be dealing with.”
The app also gives advice — both medical and practical: “If we tell you you’ve got tuberculosis, clearly there’s things that you’ve got to do medically. But also there are things that you need to do right now. Helfie seeks to address all of that.”
A big factor which makes this approach potentially so powerful is that it opens up the possibility of taking a more holistic view of an individual’s health, by connecting the dots between different conditions they might have in a way that is not typically done in traditional healthcare delivery, which is very focused on diagnostic pathways, and in which health data is often siloed.
“Our model looks at vertical conditions and different measures of what your body’s doing and uses a semantic AI infrastructure to start joining together dots for you. And it’s not just limited to what your physiology is doing. It also looks at environmental and lifestyle factors, such as what you’re eating, where you live. All that factors into the way the algorithm talks to you about your health and what it’s observing.
“A big problem in healthcare is that, despite the presence of all sorts of electronic health record systems, one of the leading causes of death is misdiagnosis by a doctor. It’s not because they’re incompetent or our education system is failing to train them. It’s because the 15 or 30 minutes they have with a patient is not pregnant with the data that they need to make accurate insights on what’s going on with that person. That’s a solvable problem. We have enough data to solve all sorts of problems, but it’s stuck in silos. It’s gatekept. It’s hoarded by all sorts of actors.”
The app is designed to allow the user to incorporate as many different data sources as they want, making it more and more personalised as they do so:
“It’s powered by your own data, which you are the owner of. It’s completely personalised. You can ask exactly the same question on blood pressure, for example, and get a different response to the next person, because you have a different age, different lifestyle etc.
“In most modern countries, the health system will give you your health record data if you ask for it. You can import all of your wearable data as well, if you’re rich enough to have a wearable. You can even put your DNA report in, and the algorithm will calibrate its risk for you based on that.
“When we talk about a personal health AI, the idea is that it is all-encompassing. It holds and manages all your data, and puts that data to use. It considers all of your health history when making assessments or recommendations. That is the algorithm’s inherent power: leveraging that contextual data about your health, based on what you choose to tell it.”
Tomeski says the aim of all this is to change the underlying economic infrastructure and incentive structure in healthcare, at least around early detection and prevention, and build a new system geared towards optimising individuals’ health algorithmically.
“We can treat pretty much everything if we catch it early enough. We’re trying to get to a point where early detection and preventative health is a live truth for all of us, not just a slogan that governments or health bodies just bandy about.
“One of the big problems that we’re looking to tackle here is the lack of participation in the prevention side of healthcare. About 70% of people succumb to a condition that could have been prevented, such as cardiovascular or respiratory conditions. It was the same figure 40 years ago. We’ve made no dent despite all the money we spend on health and all the medical marvels that come at us daily these days. In countries like the UK, the US and all the other countries that spend big bucks on health, life expectancy has been declining for the past 20 or 30 years, despite record investment in health.
“No one should succumb to or suffer from a condition that we know how to identify and treat. The fact that they are is a function of how difficult it is to operate in healthcare, how expensive it is, and all the structural impediments to improving healthcare delivery.
“It’s also a function of convenience: humans are lazy; we justify away risk; we ignore stuff. So our whole premise was: how do we make this so convenient that it becomes a case of just pointing your phone at a problem? And how do we make this so low cost that the price point is negligible in terms of the value that it generates? And how do we make sure that the data we’re generating is in active conversation with your health, for your health? And how do we make sure that the data we’re generating feeds the health system? Our view was let’s just bring that data to the edge, next to each human, where it should be, and activate it for each human. If we do that, the data becomes profoundly powerful for your own health and wellbeing.”
Helfie has raised around $35–40 million to date. But Tomeski says the company is currently embarking on a larger round, in which it is looking to raise around $250 million. “We’ve got some early commitments for about half of that already, before the official process has started. It might include venture capital investors, it might be family offices, impact investors, private equity, and possibly some sovereign wealth funds,” he says.
“When I talk to investors and partners, I tell them we’re looking to build a system for early detection and prevention that’s scalable across populations, that’s low cost, and that covers all or almost all health conditions. That doesn’t exist today. No one is trying to build what we are building, at the scale that we’re doing it at, nor with the capability or AI stack that we’ve built. No one’s attacking the problem the way we are. They’re trying to replumb the existing system, which I just don’t think is going to cut it.”
Tomeski says the potential impact of the capability Helfie is building is profound:
“Helfie AI is the world’s first operating system for human health, scalable across all eight billion people. We are working with governments and enterprises across the world to bring preventative health to entire populations. If all eight billion people on Earth, along with their health data, could actively participate in (preventative) healthcare, we would fundamentally alter the trajectory of human life. Helfie was built to make this possible — by engineering a planetary-scale health network that senses, understands, and acts before illness takes hold.”
0 comments